Research & Innovation
SsgCaps Rebuilds a Sound Scene Dataset From Public-Domain Audio So Everyone Can Use It
Sound Scene Generation — automatically synthesising artificial sound scenes — needs shared evaluation data. SsgCaps, posted September 22, is a publicly available dataset of human-engineered sound scenes, each matching a precisely structured prompt drawn from an action-based typology. It rebuilds the unpublished DCASE 2024 Task 7 reference set using only public-domain audio.